Patterns or Claims: Do they help in communicating design advice?

advertisement
Patterns or Claims: Do they help
in communicating design advice?
Michael E. Atwood
George Abraham
College of Information Science and Technology College of Information Science and Technology
Drexel University
Drexel University
Philadelphia, PA USA
Philadelphia, PA USA
atwood@drexel.edu
george.abraham@ischool.drexel.edu
and sharing. Patterns or claims can also possibly serve as
middle ground between high level guidelines or
principles and concrete instances of design.
ABSTRACT
It has been asserted that patterns or claims will help
capture and communicate interaction-design advice. Both
structures attempt to provide advice in context along with
the justifications for fit. These properties aim to make
patterns or claims more concrete and comprehensible to
novice designers than design guidelines. However,
empirical work evaluating these promises is lacking. This
research presents a controlled study that examines the
value of structuring design advice as patterns or as
claims. Patterns and claims seem different given their
respective roots in architecture and design rationale. They
also differ in their emphasis when capturing design
decisions; patterns emphasize capturing a problemsolution pair in a certain context, whereas claims focus on
capturing the positive and negative implications to a
design decision. The findings from the study suggest it
may be promising to combine the claim and pattern
structures and that such a structure may facilitate
discussions of design trade-offs.
Patterns and claims share at least two goals in common:
(1) reuse and (2) communicating design advice. It is
unclear, however, whether patterns or claims actually
add value to the design process. At present, there is no
clear empirical evidence demonstrating the benefits of
capturing advice as pattern or claims. This study corrects
this deficiency by exploring whether expressing design
advice using the pattern or claim-structure influences its
perceived effectiveness.
This study differs from past research in two ways:


Author Keywords
It evaluates possible benefits of structuring
design advice in a pattern or claim form rather
than the efficacy of using an existing pattern
collections or claims library
It studies what impact such a structure may have
on sharing design advice.
By focusing on structure of advice, we aim to explore the
benefits and challenges surrounding communication and
reusability of design advice. This will enable us to extend
the current discussion on patterns or claims by specifying
how we might articulate design advice.
Pattern, claim, controlled-study, context, trade-offs,
advice
ACM Classification Keywords
H5.m. Information interfaces and presentation (e.g.,
HCI): Miscellaneous.
This paper is organized as follows. First,we provide a
brief introduction to patterns and claims and their
inherent structure. After this we offer an overview of
empirical work examining the value of patterns or claims
to communication. A comparison of pattern and claim
structure is then shown to hghlight the affinities and
differences between the two approaches. We then
describe the methodology used to compare the two
structures followed by implications of our findings to
future work in this area.
INTRODUCTION
How should HCI researchers and practitioners
communicate design advice to their clients, end-users, or
to each other? For a discipline with a strong belief in an
iterative design-evaluate cycle, how advice gleaned from
one evaluation cycle feeds into the next design cycle is
clearly important.
It has been argued that HCI design knowledge is
traditionally presented in a format that is comprehensible
only to members of the HCI community [32]. Given the
inter-disciplinary nature of design, sharing design
knowledge may be critical. Some have theorized that the
concrete and contextual nature of advice presented as
patterns or claims will facilitate comprehension [15] [32],
PATTERNS AND CLAIMS IN HCI
This section provides an introduction to patterns and
claims to describe how their structures have been
described in the literature.
Patterns in HCI
Patterns in HCI draw their inspiration from patterns in
architecture, introduced by Alexander et al. [1]. The HCI
community has proposed that interaction patterns may
serve as a common language for users and designers to
voice their ideas and as a possible format to communicate
OZCHI 2009, November 23-27, 2009, Melbourne, Australia.
Copyright the author(s) and CHISIG
Additional copies are available at the ACM Digital Library
(http://portal.acm.org/dl.cfm) or ordered from the CHISIG secretary
(secretary@chisig.org)
OZCHI 2009 Proceedings ISBN: x-xxxxx-xxx-x
1
interaction design knowledge to those interested in HCI
[3] [5] [15].
However, there is no universally accepted definition for a
pattern in HCI; different researchers use different
definitions. Therefore, we attempt to stay as close as
possible to the definition proposed by Alexander et al [1].
The pattern structure brings together three elements- a
description of the recurring design problem, a solution
showing a configuration that brings into balance the
“forces,” and a definition of the context in which the
problem-solution pair holds true. For example, an
ALCOVE pattern describes a spatial configuration
(solution) that balances opposing forces of community
and privacy (problem) in large communal rooms
(context).
Claims in HCI
Claims analysis involves identifying the positive and
negative implications on usability, or use, stemming from
a design choice. A scenario of use involving the task and
artifact forms the basis for this reflective exercise. Carroll
and Kellogg [8] argue that HCI artifacts present
opportunities for building theory in HCI. Such efforts
may serve as a middle ground between concrete instances
of use and abstract theories in HCI. They assert that “HCI
artifacts embody psychological claims about contexts-ofuse” [id, p.8]. Claims analysis highlights design trade-offs
when making a design choice. Carroll and Rosson [9]
describe the basic claim structure as follows:
“(Artifact feature or technique) CAUSES (desirable
psychological consequence) BUT MAY ALSO CAUSE
(undesirable psychological consequence)” [id., p.193].
For example, a flashing window icon on the task bar of a
computer will alert user that a background process needs
attention, but may also distract the user from the current
window or task (it is possible for a feature to have more
than one positive or negative consequence).
To be accurate, a claim is an evalautive statement
expressed as a positive or negative design implication.
For the sake of clarity and convenience, we shed the term
claims-analysis. Instead, we adopt claim-advice to refer
to design advice that is presented as a usage scenario
along with possible pros and cons (claims) of
implementing a design decision.
EMPIRICAL WORK ON PATTERNS AND CLAIMS
A frequent argument offered for both patterns and claims
is that they are helpful in communicating design advice.
Before evaluating this promise, we review related
research in the section below.
Dearden and Finlay's review of patterns and pattern
languages in HCI reveal clearly that there is insufficient
evidence demonstrating benefits of patterns or pattern
languages, either to the design process or the designed
product [14]. We have seen several recent studies
exploring this question in the area of ubiquitous
computing [11] [29], participatory design [12] [13],
teaching HCI [4] [22], and information retrieval [34].
Among them, the controlled studies [11], [22], [29], [34]
focused on the efficacy of using patterns for improving
the quality of the design product. The evidence from
these studies is not conclusive as to whether patterns help.
In fact, these studies show no clear advantage of using
patterns to design interactive systems. Nevertheless, these
studies all argue that patterns may help communication
during the design process, and should be explored further.
Studies discussing the direct benefits of reusing claims to
design are similarly lacking. A survey of empirical work
on claims can be broadly classified under three
categories: discussions on the theory behind claims and
its application [7] [9]; using claims analysis as part of the
scenario-based design (SBD) framework [10] [28]; and
how to organize claim library to promote reuse of claims .
In terms of value to the design process, Carroll and
Rosson [9] point to projects where a scenario-based
claims analysis was applied with success. A detailed
discussion on using claims analysis as part of teaching
HCI using case studies and SBD is also available [30].
These studies, however, do not to evaluate how claims
communicate design knowledge in situations when claims
are reviewed outside the context from which they were
derived. Nor do they show clear empirical support for the
value of claims reuse in design. That is, how an archived
claim will help a designer with a design problem at hand.
There are studies reported that evaluate findability of
claim in a claims library. These studies seem targeted at
understanding how to organize claims to promote access
or findability [16] [17] [26], and not on assessing the
usefulness of advice presented in or as claims.
In summary, we did not find much evidence of controlled
studies involving claims reuse or patterns that evaluate
the impact on communicating design advice. Discussion
have mainly focused on the form and organization
[6][27]. We believe improving access to design advice is
important, but it's equally, if not more, important to
examine whether the design advice is perceived as usable
and effective first. Without evidence showing that
articulating design advice as patterns or claims facilitates
sharing and comprehension of design knowledge, the
value of organization schemes, however well-designed
and argued, is unclear.
From what we have seen, discussions of patterns or
claims seem united in purpose, but except for a brief
comparison between patterns and claims in [14], research
on these two approaches rarely intersect. We too see
patterns and claims share a lot in common. Both:
▪
▪
▪
Emphasize delivering design advice situated in
the context-of-use
Share a common goal of abstracting from
successful solutions for design problems, and
extending HCI design knowledge.
Are aimed at facilitating communication of
design advice by making design advice usable.
Discuss design trade-offs
COMPARING PATTERNS AND CLAIMS
Patterns and claims are arguably similar in many respects.
Both describe the context in which the design decision
will be deployed. Both describe the solution that this
decision implements. They differ in that while patterns
explicitly address the problem to be solved, claims do
not. Rather, claims focus on the advantages and
disadvantages of the solution while patterns consider only
the advantages.
There are three areas in which patterns and claims differ
– (1) in whether they are problem-driven (patterns) or
solution-driven (claims), (2) how exactly they define
context and (3) how explicitly they present the rationale
for design decisions.
Problem-driven or Solution-driven
Pattern descriptions tend to include a specific problem
statement that highlights the constraints present in the
design space [14]. These constraints stem both from the
task (e.g., shopping for furniture on the web) and the task
context (i.e., aspects unique to furniture shopping, such as
coordinating colors or size). However, we have seen
patterns in HCI replace this problem statement with a
generic task description, which states what the user is
attempting to do. For example, the problem solved by an
accordion menu is that an user needs to find an item in
the main navigation.
A claim usually does not include an explicit problem
statement [14]. Rather, a usage scenario motivates a claim
[31]. For example, in discussing reuse of claims such as
“virus warning through informational pop-ups” [16] or
“rare event monitor” [31], a usage scenario sets the
context for discussing pros and cons of adopting this
solution. The object of claim-advice seems to be
presenting a balanced view for a particular solution given
a usage context.
In this sense, patterns can be argued as problem-centric
versus claims, which seem solution-centric. This
distinction may be important considering that when
designers are looking for advice, its usually to find a way
out of a problem for which they don’t have a solution. So
describing the problem explicitly may enable a
comparison of the problem at hand with that described in
the pattern. Expressing trade-offs as part of the solution
may be equally important as this will enable designers to
evaluate the solution itself.
Definition of Context
Solutions in patterns are abstracted from multiple
instances of a common problem. Patterns attempt to
include a range of contexts where the problem-solution
pair holds true. There is no set rule about how many
solutions or situations need to be considered before
declaring it a pattern. Reviewing patterns in HCI will
show an explicit discussion of context under a “use
when” section describing some leading aspects of
situations where this solution is relevant. However, we
did not find any recommendations on how context needs
to be articulated, nor did we find any clear strategies used
in discussing context.
Claims do not explicity focus on describing context. A
usage scenario (task-artifact) situates a claim, without
which a claim loses its concreteness and context. In terms
of scoping the context-of-use, claims do not attempt to be
exhaustive, but instead resort to considering the leading
aspects of the usage scenario and how the intended
solution will affect the user.
Conveying Design Rationale
Pattern descriptions include justifications as to why it is a
good solution. The supporting rationale tends to be
sourced empirically (i.e., the solutions need to exist).
While most pattern descriptions include rationale, it is not
clear what or how it should be recorded. The core pattern
structure does not require an emphasis on negative
consequences of implementing this solution or
limitations. Example instances of the solution (i.e.,
screenshots) are included along with the advice to
reinforce that these solutions indeed exist.
The rationale aspect in claims seems more fleshed out
than that in patterns, and more detailed instructions
describe what to capture [9]. Unlike patterns, claims
describe both the positive and negative consequences on
the use or usability. The basis for claims can be analytical
or empirical. Furthermore, claims tend to present a
verified solution rather than generalizing from multiple
successful solutions.
Patterns and Claims: Review Summary
Given what we now know about patterns and claims, the
contribution of this research lies in exploring how these
similarities and differences contribute to the effectiveness
of design advice. Rather than examining whether
participants in our study produced patterns or claims, we
looked at how these structures help explainand
communicate interface design decisions. .
In summary, this study explores the following two
research questions: (1) What impact does the advice
structure used have on communicating interaction design
advice? (2) How does advice structure impact the time
required to prepare design advice?
METHOD
To evaluate whether the advice structure impacts the
effectiveness of design advice, a between-subjects design
was used to randomly assign 57 participants into three
groups: pattern-derived structure, claim-derived structure,
and control (no specified structure). All participants were
undergraduate students from an information and
technology school, and had completed coursework in
HCI. Participants were familiar with the concept of easeof-use, design principles, and heuristics (completing at
least an introductory course in HCI was a pre-requisite
for the study).
All participants were asked to adopt the role of a designer
and recommend a menu design for a room planner
interface. Participants received a simple version of a
room planner (Figure 1). The instructions stated that a
room planner will enable users to add and remove
furniture from an on-line catalog to a room plan, and
allow them to try out different arrangements.
To focus the participants on capturing design advice,
instead of asking each of the participants to come up with
their own menu design, they received a sheet with a
randomized ordering of three menu options: Directorytree, Drop-down, and Accordion. These options were
selected because they are fairly common in interactive
systems. These options were presented side-by-side and
included a brief set of features describing the menu. For
example, the Directory-tree menu listed the following:

Expand/Collapse different sections using the +/-

Multiple sections can be opened at a time

Product information and image displayed inside
each menu

Fixed height for the menu

Scroll (outside) to view the open catalog.
The other two menus also listed five features customized
for the menu type (i.e., for the Accordion menu it read
that clicking on the arrows open/closes the section).
Participants were asked to rank order the three menu
options and then to explain their top pick. Ranking was
introduced to prompt participants to consider all the three
options. Depending on the assigned condition,
participants were given additional instructions;
participants in the pattern and claims condition were
asked to address three questions when explaining their
menu choice (see Table 1).
All participants were asked to explain their design
decision as if they were teaching someone else to make a
similar choice, and such that others reviewing this design
advice may not see the menu options they rejected. Their
recommendations were rated for quality by two expert
judges (each having over 20 years experience in HCI
education and research; an approach adopted by other
controlled studies on patterns). No absolute measures
exist to evaluate design advice because no representation
is complete or sufficient in itself to fit all purposes;
nonetheless, the following four criteria were adapted from
other studies [20] [24] [25]:
▪
Context of Use Described or how well does the
design advice describes the context of design
advice (i.e., usage scenario, specific examples)
▪ Rationale Expressed or whether there were
justifications present for calling the menu option
a good fit
▪ Usefulness or how well does the design advice
address the fit between the design choice and the
problem context (i.e., relevancy of rationale)
▪ Overall Quality or in general how good is the
prepared design advice
In addition to these four measures, time taken for
completing the exercise was considered as an additional
dependent measure.
The recommendation or design advice was rated on these
four criteria using a 7-point Lickert-type scale. For each
recommendation, the ratings from the two judges were
averaged. Treating the rating data as interval-type,
intraclass correlation coefficient (ICC) was used to assess
reliability of ratings between the two judges.
Figure 1. Participants recommend one menu option, from
the following three, to be used on the room-planner
interface for users to peruse the furniture catalog and make
their selections: a) Directory-Tree; b) Drop-Down; c)
Accordion.
Table 1. Additional instructions provided to the participants
based on the assigned condition.
Pattern Condition
What is the user trying to do?
How will the recommended menu option be used?
Explain why it's the best option in this given situation?
Claims Condition
How will the recommended menu option be used?
Advantages of using this design choice?
Disadvantages of using this design choice?
FINDINGS
Inter-rater reliability (ICC) was 0.81 overall, which
showed a good level of consistency.
Of all the dependent measures (context, rationale,
usefulness, overall quality, and time), analysis of variance
(ANOVA) of the rating data found the differences in
usefulness to be significant, F (2, 18) =6.783, p < 0.0125.
The p value was set at .0125 level as part of
Boneferroni’s correction (Fields, 2005) to account for
type-II error. A post-hoc analysis using Scheffe’s posthoc criterion showed that the advice prepared using the
claim structure received higher mean ratings than either
the patterns or control groups. Table 2 below shows the F
and p values from ANOVA for the other variables found
not significant.
The advice structure did not affect the time required to
prepare recommendation. The overall quality between
the three was also not significant.
Measure
F Value
p Value
Context
F (2, 54) = .52
p = .596
Rationale
F (2, 54) = .859
p = .429
Quality
F ( 2, 54) = .74
p = .48
by discussing advantages and disadvantages of their
design choice, the participants in the claim condition also
considered a significantly higher number of design
implications. The number of unique design issues had a
significant and strong correlation with usefulness
(r=.686), rationale (r=.559), and overall quality (r=.598)
ratings, at p < .05 level, which seem to corroborate what
we discovered from expert judging. Context ratings were
weakly correlated with number of unique design issues
(r=.224) and was not significant either.
Time
F (2, 54)= .76
p= .44
Rationale Presented in Design Advice
Table 2. Results from ANOVA for each variable
Table 3. Descriptive statistics showing means(M) and std.
deviations (SD); * significant at p<0.05. A 7-pt Lickert type
scale was used.
Results in (M, SD) format
Criteria
Pattern
Claim
Control
Context
3.00, 1.07
2.71, 1.13
2.68, 0.96
Rationale
3.50, 0.88
3.84, 1.29
3.39, 1.08
Usefulness*
3.13, 0.99
3.95, 1.12
2.79, 0.85
Quality
3.18, 0.87
3.34, 0.90
3.00, 0.83
Time (mins)
12.73, 5.14
13.15, 5.07
11.14, 5.63
Usefulness of Advice
The advice structure had an impact on the usefulness of
design advice. Claim structure performed significantly
better than the pattern or control. Advice was rated more
useful based on how many relevant implications were
considered. To distinguish usefulness from rationale,
rationale assessed whether reasons were given for picking
a particular menu (evaluative comments) whereas
usefulness evaluated the content of the cited
implications.
Compared to the pattern and control condition, claim
advice contained a significantly higher number of unique
design implications, F (2, 54)= 6.32, p=.003. If we look at
the advice structures provided to the participants, pattern
structure differed from the claim structure in two ways.
First, it asked participants to discuss “what the user was
trying to do”. Secondly, the claim structure specifically
asked participants to consider the disadvantages of their
design choice. Pattern condition asked participants to
describe why the suggested menu was appropriate in the
given situation. Participants in the control condition did
not specify any set structure, and were asked asked to
explain why they would recommend a menu option.
The claim advice seems to have benefitted from a
discussion of disadvantages. A subsequent contentanalysis of prepared advice grouped the evaluative
comments under ten types of design implications that
were considered by the participants (e.g., visibility,
efficiency). These represented a set of design issues
considered by the participants as a whole. We found that
Expert ratings did not reveal a significant difference
between the three conditions when expressing rationale.
In other words, the structure did not impact how much
rationale was provided. In examining how participants
argued for their design decision, we found that the claim
condition elicited significantly higher number of
evaluative comments (positive+negative) compared to the
control condition or pattern condition, F(2, 54) = 6.37, p
= .002. An example positive evaluative comment was
stated as, “This design choice will allow the user to know
all of their options at any one point in time (multiple
categories can be open at once). The user can scroll
through the entire open category at once.” The pattern
and control condition almost rarely discussed negative
implications of design choice. The difference in the
number of evaluative negative comments was was
significant, F (2, 54) = 114.882, p=.001.
We observed that the claim structure prompted
participants to be more reflective about their
recommendation, both resulting in a higher number of
evaluative comments, and as we saw earlier, lead to
consideration of more design implications. This leads us
to theorize that it may help to make explicit negative
implications of design decisions in design advice.
Context of Use in Advice
The difference in the mean ratings for context of use
based on advice structure was not significant. And while
the context-of-use ratings was found to be higher for the
pattern condition, this margin was not significant.
Examining the content of design advice showed that
participants in both pattern and claim condition discussed
their design more than did the control group (c.f. table 4
on word-count). The difference in mean word count was
significant between patterns-control and claims-control, F
(2, 54)= 4.27. p=.02.
The content of advice when compared on aspects that
may be commonly associated with context (e.g., user,
task) showed that pattern advice discussed context related
aspects more than the control group. The number of
words describing context also showed a strong correlated
with context-of-use expert ratings (r=.509, p=.001).
Table 4. Means and Standard Deviation for #Words used in
the three conditions.
Condition
Mean
Std.
N
Deviation
Pattern
200.3684
78.09838
19
Claims
200.4211
72.79752
19
Control
143.1579
56.26649
19
Total
181.3158
73.63791
57
DISCUSSION
Controlled studies in the past explored whether exposure
to a set of patterns impacted the resulting quality of the
interface [11][29][34]. The results from these studies
were not conclusive in favor of patterns. Nevertheless, it
was argued that patterns may help communicate design
knowledge; and therein lies the real value of patterns. A
similar argument can be found in the claims-reuse
literature where claims were promoted as a usable and
coomprehensible form of interaction design advice [32].
This controlled study explored the promise of
communication by examining whether presenting design
advice using a pattern-derived or claim-derived structure
influenced the effectiveness of advice. A pattern-derived
stucture suggested a problem-solution-context form
whereas the claim-derived structure suggested
participants to document advice as a usage scenario, pros
and cons of the design decision.
In the following discussion we discuss our findings with
respect to usefulness, rationale, and context of use, and
what these findings mean for pattern or claim research for
sharing design advice.
Case Against “Natural” Design Advice
It appears that the pattern structure, as presented, did not
offer a strategy for participants to easily identify and
explore possible design implications. Given the freedom,
participants displayed a tendency to “sell” their idea, and
in the process were less critical. The control condition
also displayed a similar tendency.
Offering only positive evidence, in the pattern and the
control-advice, can be argued as being a “natural”
strategy for communicating design advice. The findings,
however, show that “natural” may be not be enough. The
pros and cons approach in claim structure seems to have
helped participants in considering more number of design
implications, and in the process considered more design
issues. This forces us to reconsider the value of
discussing cons in design advice.
Despite the argued affinities between design in HCI and
in architecture, pattern-advice for interactive systems may
require a different treatment. Alexander’s [1] version of
patterns does not explicitly address negative implications.
Does that imply interaction design can afford to do the
same? We can think of at least two reasons why this is the
case for architecture: (1) we may be implcitly aware of
the cons or residual forces, through our lived-in
experiences, and pointing them out may be redundant; (2)
A pattern contains a solution that helps bring the forces
into balance, such the net forces are zero, which may be
interpreted as no negative implications.
Lived-in Experiences and Interactive Systems
We believe cons, or negative implications, may prove to
more helpful in advice about interactive systems since we
do not have the benefit of lived-in experience. And hence,
may be unaware of the negative implications.
Furthermore, it’s highly unlikely that any design solution
is perfect enough such that there are no negative
implications. Patterns in collectons like Tidwell’s [33]
collection seem to acknowledge this, and try to address
negative implications in the pattern description. However,
the impact of describing negative implications has not
garnered much attention in the pattern litereature. Even
the standardized format suggested for formating patetrns,
pattern language mark-up language (PLML), while being
extensive, does not acknowledge the possible value of
cons to communicating design advice. We feels negative
implications may be just as important to communicating
design advice, as is describing context or sharing a
“good” solution.
Sufficiency of Context in Advice
The challenge for capturing context is to balance
specificity of advice with generalizability. While an
exhaustive specification of context may lend concreteness
to design advice, it may also impede the designers’ ability
to critique this advice in reuse situations. Moreover, an
exhaustive specification is impossible. One primary
argument favoring pattern or claims has been the
presence of context in design advice. In the study,
pattern-advice and claim-advice did not differ much when
discussing aspect related to context.
It’s even more challenging to articulate context separate
from the design solution (or problem) itself. Orginally,
patterns were about a recurring problem and a good
solution discussed in context. However, over the years
pattern collections have inadvertently pushed an agenda
for discussing problem, solution, and context. The
evidence of this lies in the inclusion of context in “usewhen” or “why” sections in pattern descriptions. In doing
so, it gives the impression that the problem and solution
can be abstracted from the context, and seperating the
context is a good thing. The design advice prepared by
the participants showed that it was more natural to pair
each design implication with the context of use. In fact,
the design implication lost its meaning without specifying
the conext of use alongside.
We found that using a pros and cons approach prompted
participants to implicitly describe leading aspects of
context. For example in stating a positive implication the
participant states, “the +/- button on a directory-tree menu
will be familiar to users' based on their prior experience
to indicate more or less items.” In offering such a positive
evaluative comment, the participant implicitly touched
upon aspects of the interaction context like users’
experience level. The pattern structure by itself did not
provide much guidance on describing context. The
discussions on patterns in HCI too have eluded providing
any recommendations on how to approach discussing
context. This is not a critique of any particular pattern
collection, but more a reflection on the state of pattern
research. We found that drawing the attention of our
participants to explicitly account for context did not make
a significant difference. We found that the claim structure
seems to have achieved what a pattern-advice was meant
to be, describing advice in context.
CONCLUSIONS AND FUTURE WORK
Patterns and claims both present themselves as viable
options for sharing design advice, but empirical work
examining their similarities and differences is very
limited. Furthermore, discussions on patterns and claims
rarely overlap. This controlled study found that design
advice prepared using the claims structure was perceived
to be more useful.
We plan to pursue future work in following directions:
(1) exploring a new formalism for sharing design advice;
(2) expand the discussion on patterns or claim as a
complimentary approach to existing design methods,
where the emphasis is on learning from design trade-offs.
Rather than declaring pattern or claim structure as a clear
winner, we believe future work in this area should
explore a hybrid structure, a new formalism, for sharing
design advice. This new structure proposes articulating
design advice as problem and solution in context, and
then explicating the pros and cons of applying the
solution. The concept of including a problem description
is borrowed from the pattern concept. The solution can be
presented usage scenario that situates the design artifact
in use. And finally, adopt a pros and cons structure for
articulating trade-offs. Describing the problem in context
will allow readers to evaluate the solution against the
problem itself, and specifying the pros and cons of the
solution will afford readers an opportunity to evaluate the
solution itself.
More importantly, asking whether patterns or claims
communicate design advice better than guidelines
because they are concrete and include context may be the
wrong question! Patterns or claims are not just contextrich versions of guidelines, but are more about describing
trade-offs in design. This latter contribution has been
under-appreciated. Given that design problems are messy
(a set of intersecting design issues), good design is about
making the right trade-offs. We do not yet have a formal
way to learn about design from design (trade-offs). We
believe patterns and claims may offer a way to capture
and share successful design trade-offs in context.
REFERENCES
1. Alexander, C.: The timeless way of building. Oxford
University Press, New York (1979)
2. Alexander, C., Ishikawa, S., Silverstein, M. et al.: A pattern
language: Towns, buildings, construction. Oxford
University Press, New York (1977)
3. Bayle, E., Bellamy, R., Casaday, G. et al.: Putting it all
Together: Towards a Pattern Language for Interaction
Design: A CHI 97 Workshop. SIGCHI Bulletin, 30 (1998)
4. Borchers, J.: Teaching HCI Design Patterns: Experience
from Two University Courses. Minneapolis, MI (2002)
5. Borchers, J.: A pattern approach to interaction design. 1st
edn. Wiley (2001)
6. Borchers, J.: CHI Meets PLoP: An Interaction Patterns
Workshop. SIGCHI Bull., 32 (2000) 9-12
7. Carroll, J. M., & Rosson, M. B.: Deliberated Evolution:
Stalking the View Matcher in Design Space. Design
Rationale: Concepts, Techniques, and Use, (1996)
8. Carroll, J. M., Kellogg, W. A., Rosson, M. B.: The TaskArtifact Cycle. (1991) 74-102
9. Carroll, J. M., & Rosson, M. B.: Getting Around the TaskArtifact Cycle: How to make Claims and Design by
Scenario. ACM Trans.Inf.Syst., 10 (1992) 181-212
10. Chin, G., Rosson, M., Carroll, J.: Participatory Analysis:
Shared Development of Requirements from Scenarios.
(1997) 162-169
11. Chung, E., Hong, J., Lin, J. et al.: Development and
Evaluation of Emerging Design Patterns for Ubiquitous
Computing. (2004) 233-242
12. Dearden, A., Finlay, J., Allgar, E. et al.: Using Pattern
Languages in Participatory Design. Proceedings of the
Participatory Design Conference, (2002) 104-112
13. Dearden, A., Finlay, J., Allgar, L. et al.: Evaluating Pattern
Languages in Participatory Design. Conference on Human
Factors in Computing Systems, (2002) 664-665
14. Dearden, A., & Finlay, J.: Pattern Languages in HCI: A
Critical Review. Human-Computer Interaction, 21 (2006)
49-102
15. Erickson, T.: Lingua Francas for Design: Sacred Places and
Pattern Languages. (2000) 357-368
16. Fabian, A., McCrickard, D. S., Felton, D. et al.: Designing
the Claims Reuse Library: Validating Classification
Methods for Notification Systems. Proceedings of the 42nd
annual Southeast regional conference, (2004) 357-362
17. Fabian, A., Wahid, S., Bhatia, S. et al.: Creating an
Interactive Learning Environment with Reusable HCI
Knowledge. (2006) 2314-2321
18. Henninger, S., & Corrêa, V.: Software Pattern
Communities: Current Practices and Challenges. (2007)
19. Horner, J., & Atwood, M. E.: Design Rationale: The
Rationale and the Barriers. Proceedings of the 4th Nordic
conference on Human-computer interaction: changing roles,
(2006) 341-350
20. Jeffries, R.: Usability Problem Reports: Helping Evaluators
Communicate Effectively with Developers. Usability
inspection methods table of contents, (1994) 273-294
21. Karsenty, L.: An Empirical Evaluation of Design Rationale
Documents. Proceedings of the SIGCHI conference on
Human factors in computing systems: common ground,
(1996) 150-156
22. Kotze, P., Renaud, K., Biljon, J. v.: Don't do this-- Pitfalls in
using Anti-Patterns in Teaching Human- Computer
Interaction Principles. Computers & Education, 50 (2008)
979-1008
23. McCrickard, D. S., Chewar, C. M., Somervell, J. P. et al.: A
Model for Notification Systems Evaluation- Assessing User
Goals for Multitasking Activity. ACM Transactions on
Computer-Human Interaction, 10 (2003) 312-338
24. Molich, R., Jeffries, R., Dumas, J.: Making Usability
Recommendations Useful and Usable. Journal of Usability
Studies, 2 (2007) 162-179
25. Molich, R., Hornbaek, K., Scott, J.: Recommendations on
Recommendations. (2007) 1921-1924
26. Payne, C., Allgood, C., Chewar, C. et al.: Generalizing
Interface Design Knowledge: Lessons Learned from
Developing a Claims Library. Information Reuse and
Integration, 2003.IRI 2003.IEEE International Conference
on, (2003) 362-369
27. Perspectives on,H.C.I.Patterns Concepts and Tools
C.H.I.Workshop report: Perspectives on HCI Patterns:
Concepts and Tools. CHI 2003 Workshop Report. (2003)
28. Rosson, M., & Carroll, J.: Scenario-Based Design. (2003)
1032-1050
29. Saponas, S., Prabaker, M., Abowd, G. et al.: The Impact of
Pre-Patterns on the Design of Digital Home Applications.
(2006) 189-198
30. Somerveir, J., Chewar, C., McCrickard, D.: Making a Case
for HCI: Comparing Materials for Case-Based Teaching.
Frontiers in Education, 2004.FIE 2004.34th Annual, (2004)
T3H
31. Sutcliffe, A., & Carroll, J.: Designing Claims for Reuse in
Interactive Systems Design. International Journal of HumanComputer Studies, 50 (1999) 213-241
32. Sutcliffe, A.: On the Effective use and Reuse of HCI
Knowledge. ACM Trans. Comput. -Hum. Interact., 7 (2000)
197-221
33. Tidwell, J.: Designing interfaces: Patterns for effective
interaction design. O'Reilly Media, Inc (2005)
34. Wania, C. E.: Examining the Impact of an Information
Retrieval Pattern Language on the Design of Information
Retrieval Interfaces. (2008)
Download